Querix
Querix is a no-code enterprise platform for building AI assistants grounded in your own documents through retrieval-augmented generation. Published by Spanish company Dezzai, it offers cloud connectors, shared workspaces, an embeddable chat widget and a choice of language model.
What is Querix?
Querix is an enterprise platform for creating AI assistants, published by the Spanish company Dezzai SL. The legal notice describes it as a virtual agent that converses in natural language, works both reactively, answering questions about products, services and information, and proactively, pushing recommendations and updates. It is delivered as a web application and as an embeddable chat widget that a customer can drop into their own site or internal channel.
What distinguishes it, according to the vendor, is its retrieval-augmented generation architecture, which Dezzai states it owns. Rather than fine-tuning a model on company material, the assistant retrieves passages from sources the customer has approved and composes an answer from them, citing the evidence used. The vendor presents this as the mechanism that keeps hallucinations low and makes answers explainable. A second design choice follows from it: the platform is model-agnostic, and the product's own interface exposes Azure, Anthropic, Google, OpenAI, Mistral and Meta as selectable engines.
Knowledge reaches the assistant in two ways. Documents can be uploaded directly into datasets, in formats including PDF, Word, Markdown, HTML, EPUB, plain text and a range of video containers. Alternatively, ready-made connectors pull from Google Drive, OneDrive, Dropbox, SharePoint, an FTP source or a web scraper. Retrieval covers text on every tier, adds images on Pro, and audio and video on Enterprise.
Around the assistant sits a collaboration layer: shared workspaces are created per team or per project, colleagues are invited with roles and permissions, and documents and outputs live in one place. Administration includes API keys, role-based access control and audit logs.
The homepage organises the offer around five business functions, marketing, sales, human resources, customer support and internal knowledge management, and ships a library of predefined tasks across eight families. The enterprise page names vertical focus areas of legal, HR, ESG, healthcare and finance, and publishes customer stories including a Spanish city council. There is no native mobile application; the vendor states the product runs in the browser on desktop and mobile alike.
What it does
- Build a business-specific AI assistant without writing any code
- Query internal company documents in plain language and get answers linked back to their source
- Automate customer support responses around the clock and reduce escalations
- Centralise policies, playbooks and training material so new joiners can self-serve
- Produce marketing material such as ad copy, web and email copy and brand names
- Draft sales assets including pitches, call scripts, buyer personas and prospecting lists
- Speed up hiring with job descriptions, postings, interview scripts and candidate scoring
When to use Querix / When not to
A quick filter to help you decide if Querix is the right fit.
When to use Querix
- Organisations that want an assistant answering strictly from their own internal documentation, with each answer traceable to its source
- Support teams looking to deflect routine questions and cut escalations without adding headcount
- Small and mid-sized companies with no development capacity, since the builder is presented as entirely no-code
- Public bodies and administrations, the vendor's flagship reference being a Spanish city council
- Teams that want to keep control of which language model runs behind the assistant rather than accept a single imposed engine
When not to use Querix
- Anyone who needs a native mobile application: none exists on either store, and the vendor points to the browser instead
- Buyers who require a signed data processing agreement, since none is published or announced anywhere on the site
- Organisations that must contractually pin down where their data is hosted, as no country or region is ever named
- Procurement teams that demand named security certificates, because the standards displayed carry no number and no certification body
- High-volume use on the entry tiers, which are capped at 50 queries a month on the free plan and 1,000 on Starter
How to use Querix
A typical end-to-end flow, from setup to results.
- Create an account on the sign-up page; the free plan requires no payment card
- Pick the kind of assistant you need from the available types, or start from a custom one
- Create a workspace for the team or project the assistant will serve
- Build the knowledge base by uploading documents into a dataset, in PDF, Word, Markdown, HTML, text or video form
- Or connect an existing source instead, such as Google Drive, OneDrive, Dropbox, SharePoint or an FTP location
- Configure the assistant, choosing the language model and adjusting which sources it may draw on
- Test the answers and refine the dataset until the responses cite the material you expect
- Invite colleagues into the workspace and assign them roles and permissions
- Deploy the assistant, either inside the web application or through the embeddable chat widget on your own site
- For an on-premise or heavily integrated rollout, contact the vendor, since these are reserved for the Enterprise tier
Pros & Cons
Pros
- Answers are grounded in customer-approved sources and referenced back to them, which is the vendor's central design claim
- The customer chooses the language model rather than being locked to one engine
- A permanent free plan is available without a payment card
- A complete REST API with key management and public interactive documentation is genuinely available
- Cloud storage connectors are ready to use, so knowledge does not have to be re-uploaded by hand
- The publisher is a European company whose legal identity is fully published and verifiable in the Spanish commercial register
- Enterprise customers can deploy on-premise or across clouds rather than only in the vendor's environment
Cons
- There is no terms and conditions document at all: the only contractual texts are the legal notice and the privacy policy, both dated November 2024
- The security standards displayed carry no certificate number and no certification body, and the site itself attributes them to its cloud providers
- The vendor's position on training is contradictory, promising no re-training on customer knowledge while reserving the use of anonymised data to improve its AI
- No hosting country or region is ever declared, and no subprocessor list is published
- The frequently asked questions still describe an older pricing generation that contradicts the current pricing page
- No support email address is published anywhere on the site
- The impact figures shown on the homepage carry no source, no named customer and no method
Pricing & Plans
A permanent free plan is available at no cost and without a payment card, capped at 50 queries per month and 50 MB of storage. The lowest paid entry point is the Starter plan at 49 EUR per month. Pricing is quoted monthly in euros, with no annual billing option offered on the pricing page. Higher tiers are Pro at 199 EUR per month and Enterprise from 2,000 EUR, the latter quoted on request. The site does not state whether the amounts shown are exclusive or inclusive of VAT.
- Free — 0 EUR — up to 50 queries per month
- 50 MB of storage
- text retrieval
- 1 admin user
- Starter — 49 EUR per month — up to 1
- 000 queries per month
- 5 GB of storage
- text retrieval
- 1 admin user
- Pro — 199 EUR per month — up to 5
- 000 queries per month
- 10 GB of storage
- text and image retrieval
- 3 admin users
- basic analytics
- Enterprise — from 2
- 000 EUR — unlimited or custom queries and storage
- text
- image
- audio and video retrieval
- custom integrations
- white-label interface
- on-premise deployment
Data, GDPR & hosting
A consolidated view of how Querix handles your data.
GDPR overview
GDPR compliance is claimed prominently and in concrete terms. The homepage carries a GDPR compliant badge and a dedicated security page is devoted to it. The privacy policy names the controller, cites Articles 6.1(a) to 6.1(d) as legal bases, lists the six data subject rights, states that exercising them is free of charge through info@querix.chat, and points to the supervisory authority. An external Data Protection Officer is designated by name. The legal notice adds a commitment to the EU AI Regulation on transparency, explainability and absence of bias, and tells users they may be interacting with an automated system and can request human intervention. No Article 27 representative is named, which is consistent with the company being established in Spain. The reference documents are dated November 2024.
Who owns the data?
The controller is Dezzai SL, the Spanish company that publishes Querix, reachable at info@querix.chat. A named external Data Protection Officer, Professional Group Conversia S.L.U, handles privacy matters. The privacy policy places the customer in charge of the assistant's knowledge: it is the client who trains their own assistant with the files they upload into the datasets section, and that content is stated not to be shared with third parties. The same policy nevertheless reserves a right for the vendor: users must anonymise what they upload, and the vendor may then use that anonymised material to improve the performance of its AI.
Reuse rights
There is no terms and conditions document on the site, so what a customer may do with outputs is governed only by the legal notice and the privacy policy. The legal notice reserves all intellectual and industrial property in the website and its contents to Dezzai, and forbids reproducing or redistributing them commercially without authorisation; it does not, however, restrict what a customer does with the answers their own assistant generates. The privacy policy pushes responsibility the other way: customers must strip personal data from the documents they upload, must not share sensitive information, must use the system lawfully, and must review any output used to evaluate a person. Dezzai declines responsibility for results obtained when its guidance is not followed.
Data retention & training
Hosting summary
The site never states where customer data is hosted. Neither the privacy policy, nor the legal notice, nor the security page names a country, a region or a cloud provider. The only related statements are that Dezzai relies on cloud providers holding quality and information security certifications, and that its architecture is designed for minimal coupling to any single cloud vendor so that it can move between providers. Enterprise customers are offered an on-premise or cloud-agnostic tenant deployment, which shifts hosting to the customer's own environment, but the default arrangement is not described. No subprocessor list is published, and no transfer mechanism outside the European Economic Area is detailed. The marketing site itself is served from a German data centre, but that says nothing about where the platform stores customer data and is not a statement by the vendor. Anyone with a data residency requirement will need to obtain it contractually, because it cannot be established from the published material.
Things to keep in mind
Risks and trade-offs to weigh before adopting Querix.
- The security badges are not certificates: no number, no certification body, and the site itself attributes SOC 2, ISO 27001 and ENS to its cloud providers, so do not treat them as an audited guarantee
- The vendor's stance on training is genuinely contradictory across its own pages, so anyone uploading confidential material should settle the point in writing before doing so
- No hosting country or region is declared and no subprocessor list is published, which makes a data residency commitment impossible to verify from the site
- There is no terms and conditions document and no data processing agreement, so the contractual footing for a business deployment is thinner than the enterprise positioning suggests
- An assistant that answers fluently from company documents invites people to stop reading the documents; the vendor's own material stresses reviewing outputs, particularly where a person is being evaluated
- Answers are only as good as the uploaded corpus, so an outdated or partial dataset will produce confident and wrong guidance rather than visible silence
- The impact percentages shown on the homepage have no source, no named customer and no stated method, and should not be used to build a business case
Setup & Integrations
Technical difficulty
Low for a self-service start. Sign-up is open, the free plan needs no payment card, and the vendor states an assistant can be built in about five minutes using templates and a guided process, with no programming knowledge required. Connecting a cloud source runs through a standard authorisation flow. Embedding the chat widget in an existing website calls for light technical work. Effort rises sharply at the top of the range: on-premise deployment and custom integrations are Enterprise features handled with the vendor's team, and the vendor notes that rolling the agent out means training staff to work with it.
Deployment
Integrations
Supported languages
Behind Querix
Resources
All the official URLs gathered for verification and reference.
Frequently asked questions
Who is behind Querix?
Is there a free plan, and what does the paid entry point cost?
Do I need to know how to code?
What can I feed the assistant with?
Can I choose which AI model runs behind the assistant?
Is there an API?
Is there a mobile app?
Is Querix certified to ISO 27001 or SOC 2?
Will my data be used to train AI models?
Who do I contact about privacy and GDPR?
Should you pick Querix?
Querix is a real, operating product with a verifiable technical footprint: a live REST API of 118 endpoints with public interactive documentation, a working web application, and a publisher whose legal identity is fully disclosed and confirmed in the Spanish commercial register. Its positioning is coherent and reasonably unusual. Rather than promising a smarter model, it promises a traceable answer, retrieved from sources the customer controls and referenced back to them, and it leaves the choice of language model to the buyer instead of imposing one. For an organisation whose problem is that its own documentation is unreachable, that is the right shape of answer.
The reservations are about paperwork rather than product. There is no terms and conditions document at all, which means every contractual fact rests on a legal notice and a privacy policy last updated in November 2024. The security standards on display carry no certificate number and no certification body, and the site itself assigns SOC 2, ISO 27001 and ENS to its cloud providers rather than claiming them; a buyer should read those badges as aspiration, not attestation. The vendor also contradicts itself on training, promising no re-training on customer knowledge in one place while reserving the use of anonymised data to improve its AI in another. No hosting country is named, no subprocessor list is published, and no data processing agreement is offered.
Dezzai SL was incorporated in September 2024 and is a wholly owned subsidiary of Medlab Media Group SL, so this is a young company with a parent behind it. Pricing is legible up to 199 EUR per month and opaque beyond it. Worth trialling on the free tier, worth a careful legal review before anything sensitive goes in.
- Choosing a selection results in a full page refresh.
- Opens in a new window.